The People Who Know and the People Who Decide

In most companies the people with authority over AI often know the least about it, and the people who understand it have none. Fixing that gap is the actual task today.

Last week I were in a meeting where one of our senior executives, my own boss, told a story I did not expect.

He had been trying to set up his Garmin to do something specific, and he had gotten it working by talking to one of our own people, my colleague who lives and breathes this stuff, about connecting it through MCP. If you do not know what MCP is, that is sort of the point. I suppose most people don’t yet. But that should not stop us, my boss had just sat with someone who did, asked questions, and followed it far enough to get the thing working. And his message to the room was not “look what I did.” It was “pay attention to this, because I did not understand it a week ago and now I can see why it matters.”

That is the exact opposite of how most executives handle AI, and it lines up almost perfectly with something, one of my favorite authors, Simon Sinek said recently.

Sinek was speaking at SHRM26 in Orlando, and he went after the way company leaders are talking about AI. The narrative that CEOs are laying off thousands of people because AI can do their jobs, he said, is absolute nonsense. AI is not in a position to replace thousands of people in your company. And the sharper part: some leaders are creating a narrative that generates anxiety and makes AI deeply unpopular with the very workforce that has to use it. He called it short-term thinking.

He is right, and I want to start with why (pun totally intended), because I come at this from a slightly odd angle for someone who writes as much cautionary AI stuff as I do.

I am an optimist about AI. Genuinely. My posts do not always read that way, because I spend lot of my attention on the risks and the failures and the things we are not thinking about carefully enough. But that is because the optimism does not need defending and the caution does. So let me defend it, once. I think we are on the edge of something great. What these tools do today is the worst they will ever do, and almost every shortcoming I run into, the hallucinations, the context limits, the brittleness, is an engineering problem that people are solving faster than anything I have watched in twenty years. I am not nervous about the technology in the long run. I am excited about it.

Which is exactly why the leadership behaviour Sinek is describing gets under my skin. It is not the technology making people anxious. It is the people in charge of it.

Here is what actually happens in a lot of companies. Someone works out that saying “AI” on an earnings call moves the share price. So out come the words. The AI transformation. The AI-first strategy. AI-enabled everything. A kind of corporate bingo where you win by hitting the most squares in front of investors. It works on the market for a while. Inside the building it lands completely differently. Every time a leader stands up and talks about AI in big sweeping transformational terms, a lot of the workforce hears one thing, which is: they are working out how to replace me. Same words, two audiences, and leadership is usually optimising for the investors while being deaf to the people who actually have to use the tools.

You end up with the worst of both. The market gets promises that will be hard to keep. The staff get fear that makes them dig in against the very things that would help them. And the actual opportunity, the grounded practical use of AI on real problems, gets lost in the noise between the hype and the dread.

But there is something underneath all this that I think is the real problem.

In many organisations right now there is a near-perfect inversion between who knows about AI and who gets to decide about it. The people with the authority, the senior executives, the board, the ones holding the budget, are very often the people who understand it least. They have read the headlines and sat through the vendor pitches and formed opinions, but they have rarely used the tools deeply enough to know what they are actually good and bad at. And the people who do understand, the engineers using it every day, the analysts who have worked out where it helps and where it falls over, the individual contributors quietly wiring it into their own work, have almost no say in how the company adopts it. They know where the value is. Nobody asks them.

This is not new in the abstract. Authority and expertise have never lined up neatly in big companies. But AI makes the gap unusually expensive, because the technology moves too fast and is too patchy in what it can actually do for decisions made from headlines to be any good. A leader who does not understand where AI genuinely works will either over-promise, chasing the buzzwords into commitments the tools cannot honour, or under-invest, treating the whole thing as a risk to be managed. Both come from the same place. Someone too far from the actual capability is judging it.

This is what made my boss’s story land the way it did. He is senior enough to sit firmly on the authority side of that gap. And instead of doing what that side usually does, forming a view from headlines and vendor decks, he went and sat with the one person in the building who actually knew, and let that person teach him. He crossed the gap himself. He did not need to become the expert. He needed to know who the expert was and be willing to listen to him, and then say to the rest of us, out loud, that he had learned something and that keeping ourselves paying attention is extremely important right now.

That is the whole job right now, honestly. Not picking the right vendor or writing the AI strategy memo. Connecting the two halves that have drifted apart, the people with the authority to act and the people with the knowledge to act well. It does not happen on its own. It asks leaders to do something a lot of them find genuinely hard, which is to admit they are not the smartest person in the room on this, go find the people who actually get it, and give them real influence over decisions currently being made over their heads.

Sinek put the practical version of it simply. Be transparent. Give people context. Let the people who are going to use the technology advise on how to use it. That last bit is the whole thing, and my boss did exactly that on parallel. The knowledge is already sitting inside your company. It is just in the wrong place relative to the authority, and no amount of AI strategy consulting fixes that if leadership will not listen downward.

I stay an optimist, and this is part of why. The tools are going to keep getting better whether or not any single company gets its act together. But the ones that actually capture something real from it, rather than a line on an earnings call, will be the ones whose leaders close that gap between knowing and deciding. That was never a technology problem. It is a leadership problem, and leadership problems are the kind we actually know how to solve, when we are honest enough to name them.

The tools are ready to do remarkable things. The open question is whether the people in charge are ready to listen to the people who already know how to use them.